|
Abstract:
|
Understanding and effectively applying evolutionary algorithms requires predicting and explaining the evolutionary dynamics of the evolving system. However, many evolutionary systems include evolutionary forces that run counter to the designer's intent or intuition. One such force is the pressure for resilient individuals, individuals that are minimally affected by genetic changes. It is now clear that pressure for resilient (genetically robust) individuals plays an important role in determining the dynamics of evolutionary systems, both real and artificial. Evidence suggests that pressure for resiliency affects evolutionary dynamics in a number of ways, including: contributing to bloat and/or compression in variable length systems, modifying which genes are favored by evolution, and determining which peaks in a fitness landscape are accessible to an evolving population. The later implies that pressure for resiliency and its impact on evolutionary dynamics determines which solutions are likely to be found using an evolutionary algorithm. This tutorial reviews the latest resiliency results from both the biological and evolutionary computation fields and examines their practical impact on typical evolutionary algorithms.
|
